|
||||||
After a careful comparison of several methods for Seasonal Adjustment, Eurostat has decided to focus its attention on two seasonal adjustment methods: Tramo/Seats, an ARIMA model-based method, written by V. Gómez and A. Maravall and X-12-Arima of the US Census Bureau. These two methods have had to be implemented in a single interface, called "DEMETRA", in order to facilitate the application of these modern time series techniques to large-scale sets of time series and in the explicit consideration of the needs of production units in statistical institutes, especially Eurostat. Thus, DEMETRA has been developed to ease the access of non-specialists and specialists in time-series statistics to the packages Tramo/Seats and X-12-Arima and to improve largely their user-friendliness. Demetra imposes a recognised seasonal adjustment policy, and follows standard practice of seasonal adjustment.
DEMETRA contains two main modules: seasonal adjustment and trend estimation with an automated procedure (e.g. for unexperienced users or for large-scale sets of time series), and with a user-friendly procedure for detailed analysis of single time series.
The automated module allows the fully automatic treatment of lists of up to thousands of time series.
Every period (e.g. month, quarter) the system has to get the time series stored in a database, to apply to each series a given method that should be consistent with what was done before, to run the seasonal adjustment for the whole set of time series, to test if the results are satisfactory and to store the satisfactory results in the database. Thus, the interface proposes four statistical tools for a seasonal adjustment using default or customised parameters for a new automatic processing or using previous model settings including already estimated ARIMA and regression coefficients or with re-estimation of these coefficients. For doing this, DEMETRA provides features for storing the seasonal adjustment settings for each original time series in the database. These tools support a policy for reduced (e.g. only yearly) revisions of the seasonal adjustment models used. After the execution of the complete list of series, a general overview with important information on the modelling specifications and diagnostic statistics for all series is shown. For series facing some difficulties, a thorough and assisted examination is provided using up to two alternative models, more detailed diagnostic tables and graphs.
The module for detailed analysis facilitates the comparison between the results corresponding to different model specifications for single time series in practical cases.
DEMETRA includes the most important features of Tramo/Seats (April 2005) and X-12-Arima (0.3 release): pre-adjustment with automatic detection and correction of outliers, removing of calendar effects, automatic model identification/selection, forecast, seasonal adjustment and trend estimation, revision history and sliding span analysis, and diagnostic checking.
DEMETRA is a fully menu driven package running under Windows XP/NT/2000/Me/98/95. It provides access to date in FAME, MS-EXCEL, ASCII, SAS and ORACLE EXPRESS databases. (The "namelist" object of FAME can be used.)
DEMETRA was developed by EUROSTAT with participation of Jens Dossé and Servais Hoffmann, Pierre Kelsen, Christophe Planas and Raoul Depoutot.
DEMETRA 2.0
is an improvement for the look and user-friendliness. Some of the most important changes are:
completely new presentation of the "Automated Module" projects
new wizards for the creation of new projects and the SA processing
adding/removing of time series in a "Automated Module" project
printing facilities (for graphs, tables, text files and quality reports)
functions for reading, generating and translating input files in the original Tramo/Seats and X-12-Arima formats (*.inp, *.spc files)
customizable working directory (support for network-based installations), name and maximum size of log
file
AutoRecover function (automatic backup and recovery)
includes project properties e.g. for title, author, keywords and comments
context-sensitive help (will be completed with a future update of the "Demetra.hlp" file)
update of X-12-Arima to last available release version 0.2.8
DEMETRA 2.0 (Service Pack 1) - Release
besides corrections for bugs contains also new tools:
access to data in Oracle Express databases (Personal Express 5/6, Express Server 5/6)
enhanced FAME access: dramatically increased catalogue-retrieval speed, retrieves
"formula" objects, handling of usergroup access rights, better use of log-file
taylorable composite index for the sake of automatic processing
business-cycle analysis using ARIMA-model-based
calculation of the Hodrick-Prescott filter (Tramo/Seats)
automatic optimisation (reduction) of the number of
trading day regressors
(model) stability analysis for improving the yearly
modelling
expert system for a fully-automatic model for
series with rejected adjustments
more flexible and memory-saving management of projects with "very
large-scale" datasets (for Fame of Express data)
saving of calendar-effect corrected series (Tramo/Seats)
saving of parameters in original formats in the databases (for X-12-Arima and
Fame databases, the format for the X-12-Arima Fame interface is used)
possibility of saving Seats factors not as percentages but as "normal"
factors
modelling span for Tramo/Seats
smooth function for the determination of the critical value for the outlier detection (Tramo/Seats, depending on the time series length)
deletion or execution of several models at once in Detailed
analysis module
cut/copy/paste of models in Detailed analysis module and
adaptation of the model names
graphs for seasonal factors per period, and result series of
business-cycle analysis, improved selections of series to be graphed
Example of the "Automated Module" project presentation:

Example of the DEMETRA seasonal adjustment quality report:

How can one access time series in FAME databases with DEMETRA?
Scenario 1: Your FAME programme and databases are on a Windows XP/NT/2000 server:You will be able to directly access them from a connected Windows-based PC without any additional installations.
Scenario 2: Your FAME programme and databases are on a UNIX server:You will be able to access them from a connected Windows-based PC. The "Demetra Fame Server" must be previously compiled and installed on the UNIX server.
DEMETRA is freely available. Please read the legal notice while
installing the software and note that Eurostat gives no warranty for
the use of the programme. To help us further improving the
DEMETRA programme and keep you informed of further work and upgrades, we
would very appreciate your in this Seasonal Adjustment interest group or you can send messages to the Help-Desk anonymously or
self-register to the sites to get your personal login and password.
Please refer to the developers for methodological questions about Tramo/Seats or X-12-Arima.
Please refer to the Library
section of this interest group to download DEMETRA and its user
manual (If you are a new user, please give us your name, institution,
e-mail address and mail address by writing to the responsible person mentioned at the
Contact Information page
of this interst group).
This page has been created by Mr. Jens Dossé and revised by Mr. Jean-Marc Museux.
|
|
|